Citation

McMurry TL, Poplin GS. J. Biomech. 2016; 52: 187-188.

Affiliation

University of Virginia, Center for Applied Biomechanics, Charlottesville, VA 22911, USA. Electronic address: poplin@virginia.edu.

Copyright

(Copyright © 2016, Elsevier Publishing)

DOI

10.1016/j.jbiomech.2016.09.047

PMID

28034437

Abstract

We read Yoganandan et al. (2016) [doi 10.1016/j.jbiomech.2016.08.002] with interest, and are encouraged to see continued efforts to improve statistical practice in the field of biomechanics, and to continue to evaluate the ISO recommendations (ISO, 2014) for the construction of injury risk functions. We have also previously raised concerns with regard to statistical practice in this area; unfortunately the present paper does not fully address some of the issues documented in McMurry and Poplin (2015). In this note, we highlight two previously raised points of clarification plus a new point raised by these new recommendations. The first concerns the ability of relatively crude statistical techniques to “choose” optimal model formulations; the second concerns the use of confidence interval width as a quality metric; and the final is a specific concern with regard to Yoganandan et al.׳s recommendations for the use of Kolmogorov–Smirnov tests.

Language: en